MSR-IVA Enhances Multimodal Brain Data Fusion for Disease Monitoring.

Understanding how brain structure relates to dynamic functional states is crucial for monitoring disease progression. Traditional methods for fusing multimodal brain data, such as structural MRI (sMRI) and dynamic functional network connectivity (dFNC), often struggle when the same structural representation needs to be linked to multiple, changing functional states. Applying independent vector analysis (IVA) separately to each state can lead to disconnected structural decompositions, while forcing identical decompositions might obscure state-specific insights. To overcome these limitations, researchers propose Masked Structural Residual Independent Vector Analysis (MSR-IVA). This novel framework is "state-aware," meaning it intelligently combines a core, shared structural representation with specific adaptations for each functional state. It also incorporates masks to handle situations where not every subject expresses every dynamic state, ensuring more accurate and relevant data fusion. Evaluations using an Alzheimer's Disease Neuroimaging Initiative cohort demonstrated MSR-IVA's effectiveness. It significantly improved the matching of structural sources to functional states and reduced unwanted dependencies compared to baseline methods. The framework also showed controlled structural sharing, maintaining source correspondence across states while allowing for necessary state-specific adjustments, making it a powerful tool for neuroimaging analysis.